Evidence That GH115 α-Glucuronidase Activity, Which Is Required to Degrade Plant Biomass, Is Dependent on Conformational Flexibility*

Background: The structure of GH115 glucuronidases that remove glucuronic acid from xylan chains is unknown. Results: Bacteroides ovatus GH115 glucuronidase is a dimeric enzyme that contains a flexible active site pocket. Conclusion: The assembly of the catalytic apparatus of the glucuronidase requires substantial conformational changes. Significance: Conformational changes are highly unusual in glycoside hydrolases.

The microbial degradation of plant biomass in soil is critical to the carbon cycle and is of increasing importance in environmentally significant industries, particularly the bioenergy and biorefining sectors, where the plant cell wall is the key substrate (1). Furthermore, dietary plant glycans that are not degraded by mammalian intestinal enzymes are the major nutrient available to the microbial communities of the fore-stomach or rumen in ruminants, and the large bowel in monogastric mammals (2). In humans the fermentation of dietary glycans by the microbes of the large bowel (defined as the microbiota) contributes up to 10% of daily calories (3). The microbiota also has a significant impact on the host through interactions with the immune system, which in turn can affect human health (see Ref. 4 for review). Plant cell walls present a highly complex substrate, and thus microorganisms that utilize these composite structures express an array of catabolic enzymes, primarily glycoside hydrolases, but also carbohydrate esterases and polysaccharide lyases, which degrade the structural polysaccharides into their component monosaccharides (see Refs. 5 and 6 for review). These enzymes are grouped into sequence-based families in the CAZy database (glycoside hydrolase and carbohydrate esterase families are defined as GH and CE, respectively) (7).
Xylan, the major matrix polysaccharide of plant cell walls, consists of ␤1,4-linked D-xylose residues decorated at O2 and/or O3 with L-arabinofuranose and acetyl groups, and exclusively at O2 with ␣-D-glucuronic acid (GlcA) or 4-Omethyl-GlcA (MeGlcA) (Fig. 1) (8). These decorated regions make the polymer backbone recalcitrant to attack by most endo-xylanases and also block the action of xylosidases. To overcome this bottleneck, the side chains are removed by specific debranching enzymes, acetyl esterases, ␣-L-arabinofuranosidases, and ␣-glucuronidases (reviewed in Refs. 5 and 6). The majority of ␣-glucuronidases that hydrolyze the GlcA and MeGlcA decorations are located in GH67 (9,10). These enzymes, however, do not attack glucuronoxylan, they only remove uronic acid from the non-reducing end of glucuronoxylooligosaccharides (9,10). The structure of several GH67 enzymes show that O4 of the xylose decorated with GlcA is pointing at the surface of the enzymes, explaining why these ␣-glucuronidases can only target uronic acids attached to the non-reducing end of xylan chains (10,11). Recently, two microbial eukaryotic enzymes from GH115 were shown to be ␣-glucuronidases that could cleave GlcA and MeGlcA from both the non-reducing end and internal xylose units of xylan and xylo-oligosaccharides (12,13), displaying a single displacement acid base-assisted (inverting) mechanism, which results in inversion of the anomeric configuration of the cleaved uronic acid (14). The two enzymes had a higher specific activity against glucurono-xylooligosaccharides compared with glucuronoxylan, although no kinetic parameters for these enzymes were reported. As no crystal structure for a GH115 enzyme is available, the structural basis for the specificity of these ␣-glucuronidases is unknown.
To date, GH115 contains 237 members of which the majority (177) are bacterial, although only a single prokaryotic enzyme (from Streptomyces pristinaespiralis) has been shown to display ␣-glucuronidase activity (15). Bacteroidetes is a major phylum in the human gut microbiota. These organisms, exemplified by Bacteroides ovatus, are capable of degrading an extensive range of human and plant glycans (16). This glycan-degrading capacity is orchestrated by physically linked and co-regulated genes that are known as polysaccharide utilization loci (see Ref. 17 for review). The genome of B. ovatus contains seven open reading frames that are annotated as GH115s, two of which are in a polysaccharide utilization locus that is up-regulated during growth on complex xylans (16). Here we report the biochemical characterization and structure of one of these enzymes, BACOVA_03449, defined henceforth as BoAgu115A. The enzyme was shown to be a glucuronoxylan-specific ␣-glucuronidase that displayed a strong preference for uronic acids that decorate internal xylose units. The enzyme displays the same core-fold as GH67 ␣-glucuronidases, however, the active site residues are not conserved in the two enzymes. Mutagenesis and structural data indicate that a significant conformational change is required to assemble the catalytic apparatus in BoAgu115A.

Cloning, Expression, and Purification of GH115 Enzymes-
The open reading frame encoding mature forms (and truncated derivatives of BoAgu115A) of the seven B. ovatus GH115 proteins were amplified from genomic DNA (B. ovatus ATCC 8483) by PCR using primers that introduce NheI and XhoI restriction enzyme sites. The amplified DNA was cloned into the Escherichia coli expression vector pET28a (Novagen) such that the encoded recombinant proteins contain an N-terminal His 6 tag. The recombinant proteins were produced in E. coli BL21 DE3 (Novagen) cells, harboring appropriate pET-based plasmids, and cultured in LB broth containing kanamycin (50 g/ml) at 37°C. Cells were grown to mid-exponential phase (A 600 of 0.6), at which point isopropyl ␤-D-thiogalactopyranoside was added to a final concentration of 1 mM, and the cultures were incubated for a further 16 h at 16°C. The cells were harvested by centrifugation, and His 6 -tagged recombinant protein was purified from cell-free extracts by immobilized metal affinity chromatography (IMAC) using standard methodology (18). For crystallographic studies BoAgu115A was further purified by size exclusion chromatography using a Superdex 75 column. All proteins were purified to electrophoretic homogeneity as judged by SDS-PAGE.
Mutagenesis-Site-directed mutagenesis was conducted using the PCR-based QuikChange site-directed mutagenesis kit (Stratagene) according to the manufacturer's instructions, using the plasmid encoding BoAgu115A as the template and appropriate primer pairs.
Nomenclature of Oligosaccharides-The oligosaccharides are defined as follows: for homopolymers of ␤1,4-linked xylose residues either the full name of the saccharide is used or the degree of polymerization is indicated as a subscript (e.g. the tetrasaccharide of xylose is denoted as xylotetraose or X 4 ). Glucurono-xylooligosaccharides are identified by their sequence (unless otherwise stated) in which X is an undecorated xylose and U is a xylose containing an ␣1,2-linked [Me]GlcA (e.g. UXXX is xylotetraose in which the xylose at the non-reducing end contains an ␣1,2-linked [Me]GlcA).
Enzyme Assays-Substrates consisted of glucurono-xylans or glucurono-xylooligosaccharides. The glucurono-xylooligosaccharides UX and UXXX were purchased from Megazyme International (County Wicklow, Ireland), whereas UXX and XUXX were made as follows: 5 g of birch wood xylan (Sigma) was digested to completion with either CjXyn10A (19) or NpXyn11A (20) xylanases in 50 mM sodium phosphate, 12 mM citrate buffer, pH 6.5, containing 1 mg/ml of BSA (PC buffer) to completion. The glucurono-xylooligosaccharides generated by CjXyn10A, UXX, or NpXyn11A, XUXX, were purified from neutral xylooligosaccharides using Dowex chromatography (9). The structure of the products were confirmed by incubation with CjGlcA67A, which generated xylotriose from the CjXyn10A product but no undecorated xylooligosaccharide from the NpXyn11A product, whereas BoAgu115A released xylotriose and xylotetraose from glucurono-xylooligosaccharides generated by the GH10 and GH11 xylanases, respectively. The deduced structure of the two glucurono-xylooligosaccharides is entirely consistent with the glucuronoxylan binding mode of the two xylanases (21,22). Enzyme assays, in which polysaccharides or glucurono-xylooligosaccharides were the substrates, were carried out in PC buffer at 37°C using enzyme purified to electrophoretic homogeneity by immobilized metal ion affinity chromatography. The concentration of enzyme varied from 10 nM for the wild type glucuronidase to 10 M for the least active variants of BoAgu115A. For kinetic assays, glucu-FIGURE 1. Schematic of xylan. The schematic shows a basic xylan structure, the enzyme classes that hydrolyze the various linkages, and their respective major CAZy families. Simple glucuronoxylans predominate in the walls of dicots, whereas monocots usually contain more complex glucuronoarabinoxylans. ronic acid was detected using the ␣-D-glucuronidase assay kit in which the uronic acid released is oxidized to glucarate with the concomitant reduction of NAD ϩ to NADH, which was monitored continuously at 340 nm and quantified using a molar extinction coefficient of 6220 M Ϫ1 cm Ϫ1 . The molar concentration of the GlcA/MeGlcA in the glucuronoxylans was quantified by digesting 100 g of the polysaccharides to completion with BoAgu115A. To measure the activity of the BoAgu115A mutants the xylotetraose reaction product released from XUXX was monitored by high performance anion-exchange chromatography (HPAEC) as described previously (18). The reaction was carried out in 20 mM sodium phosphate buffer, pH 7.0, at a substrate concentration (1 mM) that was well below the K M . Thus, the initial rate of hydrolysis of the glucurono-xylooligosaccharides gives a direct readout of k cat /K M (19). The glucuronidase-catalyzed reactions were also subjected to Polysaccharide Analysis using Carbohydrate gel Electrophoresis (PACE) 4 as follows: a alcohol-insoluble residue was prepared from mature Arabidopsis thaliana wild type and gux1gux2 stems as well as wild type willow, barley, sugar cane, and Miscanthus stems, as previously described (23). Alcohol-insoluble residue (500 g) was pre-treated with 20 l of 4 M NaOH for 1 h, neutralized with HCl, and ammonium acetate buffer, pH 6.0, added to a final concentration of 0.1 M and a final volume of 500 l. The Arabidopsis alcohol-insoluble residue was digested to completion with xylanases (CjXyn10A and NpXyn11A), a glucuronoxylanase (BoGH30; Bacova_03432), and GH67 and GH115 ␣-glucuronidases (CjGlcA67A and BoAgu115A) as stated in the text and then dried in vacuum, whereas the other xylans were just digested with BoAgu115A. Released monoand oligosaccharides were labeled with 8-aminonaphthalene-1,3,6-trisulfonic acid (Invitrogen) and separated by gel electrophoresis as previously described (24). Xylooligosaccharides (Xyl 1-6 , Megazyme Int.) were derivatized alongside each set of samples and run as standards within each gel.
Crystallization, Data Collection, Structure Solution, and Refinement-SeMet-BoAgu115A and native BoAgu115A, at 10 mg/ml, were crystallized from 19% PEG3350, 0.2 M sodium citrate, pH 5.5. For both selenomethionine and native enzyme, crystals were harvested in a solution containing the mother liquor supplemented with 15% (v/v) PEG 400 or Paratone N oil as cryoprotectant and flash frozen in liquid nitrogen. Native crystals were soaked with 300 mM GlcA to obtain structures of BoAgu115A in complex with the uronic acid product.
Diffraction data were collected at the Diamond Light Source, Didcot, United Kingdom, on beamlines I04 -1 (ligand, ϭ 0.9163 Å) and I02 (native and selenomethionine, ϭ 0.9795 Å) at a temperature of 100 K. Data were processed and integrated with iMosflm (25) and scaled using SCALA (26). For all datasets, the space groups were determined to be P2 1 2 1 2 1 with two molecules in the asymmetric unit (giving a Matthews coefficient of about 2.7 Å 3 Da Ϫ1 and a solvent content of about 50%). Heavy atom sites were found and initial phases were calculated using the HKL2MAP interface for the ShelxC/D/E pipeline (27). All other computing used the CCP4 suite of programs (28).
Phases were extended to an initial native dataset at 2.65 Å and a starting model was built using Buccaneer (29). The model underwent recursive cycles of model building in COOT (30) and refinement in REFMAC (31). The native model was used as the search model for molecular replacement in Molrep to solve the ligand datasets (32). Solvent molecules were added using COOT and checked manually. Five percent of the observations were randomly selected for the R free set. The models were validated using Molprobity (33). The data statistics and refinement details are reported in Table 1.
Analytical Ultracentrifugation-Sedimentation velocity experiments were carried out in a Beckman Coulter (Palo Alto, CA) ProteomeLab XL-I analytical ultracentrifuge using interference optics. All analytical ultracentrifugation runs were carried out at the rotation speed of 48,000 rpm and experimental temperature of 20°C; the velocity scans were taken 1 s apart, 600 scans in total. The sample volume was 400 l. The rate protein sedimentation was used to calculate the M r of the glucuronidase as described previously (34).

RESULTS AND DISCUSSION
BoAgu115A Is a Xylan-specific ␣-Glucuronidase-The genome of B. ovatus encodes seven proteins that are members of GH115. To explore their potential enzymatic activities, six of the seven B. ovatus GH115 proteins were expressed in E. coli in soluble form (BACOVA_00249 could not be expressed in E. coli), purified by IMAC to electrophoretic homogeneity, and their activity against glucuronoxylans was explored using PACE. The data showed that BoAgu115A released glucuronic acid (GlcA) and 4-O-methyl-GlcA (MeGlcA) from wild type Arabidopsis glucuronoxylan, but not from an Arabidopsis mutant completely lacking MeGlcA and GlcA decorations on xylan (gux1gux2) (23) (Fig. 2A). The release of both GlcA and MeGlcA demonstrate that the enzyme can accommodate the 4-O-methyl substitution of GlcA, a trait shared with the other family of ␣-glucuronidases, GH67. The enzyme could also remove MeGlcA from glucuronoxylans present in bioenergy-relevant dicots such as willow, and from the more complex glucuronoarabinoxylans derived from monocots such as barley, sugar cane, and Miscanthus (Fig. 2B). Thus, BoAgu115A is a ␣-glucuronidase that targets the uronic acids that decorate xylans. None of the other B. ovatus GH115 proteins exhibited measurable catalytic activity against a range of plant polysaccharides that contain uronic acids including glucuronoxylans, homogalacturonic acid, and rhamnogalacturonan I and II (data not shown). The lack of activity of one of these B. ovatus enzymes, BACOVA_03434, against glucuronoxylan is surprising as it is closely related to BoAgu115A (54% identity), and is encoded by a gene within the same xylan-activated polysaccharide utilization locus. In contrast, BACOVA_00492, BACOVA_00982, BACOVA_02173, and BACOVA_02777 are more distantly related to BoAgu115A and are not expressed in response to xylans, which may point to specificities that are not related to the hemicellulose.
Biochemical and Biophysical Properties of BoAgu115A-To explore the catalytic properties of BoAgu115A in more detail the activity of the enzyme against glucurono-xylooligosaccharides and pre-treated glucuronoxylan was evaluated using PACE. When Arabidopsis glucuronoxylan was treated with the GH30 glucuronoxylan-specific xylanase, BACOVA_03432, a range of products were observed with the most prominent species having a degree of polymerization of 7 to 11. The GH30 enzyme was inactive against the hemicellulose that had been pre-treated with BoAgu115A (Fig. 2C), indicating that glucuronidase can remove [Me]GlcA from all locations within the hemicellulose.
Glucurono-xylooligosaccharides were generated by digesting Arabidopsis glucuronoxylan to completion (i.e. the product profile remains stable) with GH10 or GH11 xylanases, which generated UXX and XUXX as the limit products, respectively (35). The data showed that BoAgu115A hydrolyzed UXX and XUXX (Fig. 3A), indicating that the enzyme cleaves GlcA from both non-reducing terminal and internal xylose residues, consistent with its activity against xylan, whereas the GH67 ␣-glucuronidase CjGlcA67A released [Me]GlcA from UXX, but not from XUXX (data not shown), consistent with its capacity to remove the uronic acid only when it decorates the non-reducing terminal xylose (9,11).
When BoAgu115A was assayed against a mixture of terminally substituted aldouronic acids the activity of the enzyme followed the order UXXXϾUXXϾ ϾUX (Fig. 3B), demonstrat- ing that the enzyme contains at least two positive subsites downstream of the ϩ1 subsite (ϩ2R and ϩ3R). Positive subsites bind to the xylan backbone with the ϩ1 subsite housing the xylose linked to the cleaved GlcA, and subsites that bind successively to sugars toward the reducing end of the xylan chain are labeled ϩR2, ϩR3 etc., whereas the subsites that bind to xylose toward the non-reducing end are defined as ϩNR2, ϩNR3 etc. (nomenclature described in Ref. 36). It was also evident that BoAgu115A was significantly more active against XUXX than UXX ( Table 2), indicating that the glucuronidase contains at least one subsite upstream of the ϩ1 subsite (ϩ2NR). Against glucuronoxylans from birch and beech the K m of BoAgu115A is ϳ10-fold lower than XUXX, whereas the k cat of the enzyme is ϳ5-fold less against the polysaccharide compared with the oligosaccharide. Thus, although the ␣-glucuronidase has a higher catalytic efficiency against xylan than XUXX, its lower k cat likely reflects tighter binding of the deglucuronylated xylan chains (i.e. reaction product) to the positive subsites, compared with the xylooligosaccharides released from the glucurono-xylooligosaccharides, resulting in slow product release and hence turnover rate. Thus, it is possible that the xylan binding site of BoAgu115A may be able to bind more than four xylose residues.
The pH optimum of BoAgu115A was ϳ7.0 (Fig. 4A) and when subjected to analytical ultracentrifugation migrated with a sedimentation coefficient of 8.78 Ϯ 0.03 S, which equates to a molecular mass of 199 Ϯ 6.7 kDa (Fig. 4B). Given that the 824amino acid recombinant form of BoAgu115A has a molecular mass of 85 kDa, these data indicate that the glucuronidase is a dimer in solution, consistent with its crystal structure (see below).
Structure of BoAgu115A-The crystal structure of the native BoAgu115A was solved by extending the phase information from an initial selenomethionine SAD at 3.0 Å to a native res-olution of 2.65 Å, whereas the ligand structure of the enzyme was determined by molecular replacement at a resolution of 2.14 Å ( Table 1). The final model of BoAgu115A consists of one dimeric molecule in the asymmetric unit, with each protomer consisting of residues 33-856, with an average B factor of 31.5 ( Table 1). The final crystallographic R value is 20.34, with an R free of 26.59 for the native apo model, 17.40 and 21.91, respectively, for the ligand bound model. BoAgu115A is an ␣/␤ globular protein with overall dimensions of about 100 ϫ 70 ϫ 50 Å per protomer. The enzyme consists of four distinct domains, which are connected by extended loops (Fig. 5A). The N-terminal domain, residues 33-196 (all residues are identified by their position in the full-length protein), comprises six ␤-strands that lay on top of two parallel ␣-helices. The second domain (amino acids 197-482) displays a (␤/␣) 8-fold (TIM barrel). This domain deviates slightly from a canonical TIM barrel; there is an additional helix between ␤-strands ␤-1 and ␤-2, and between ␤-3 and ␤-4, and the helices protruding from ␤-5 and ␤-6 are unusually short. The two ␣-helices of the N-terminal domain make extensive contacts with the ␣-helices extending from ␤-1 and ␤-2 of the TIM-barrel of the second domain. The third domain, extending from residues 488 -641, comprises a five- In panel A, wild type (WT) Arabidopsis glucuronoxylan was treated with either a GH10 or GH11 xylanase that produce terminal or internally substituted glucurono-xylooligosaccharides, respectively, as their end products. The enzymes were inactivated by heating and then subjected to PACE before and after treatment with BoAgu115A. In panel B, a mixture of UX, UXX, and UXXX were incubated with BoAgu115A and the product generated over time was monitored by the release of the appropriate xylooligosaccharide by high performance anion-exchange chromatography with pulsed amperometric detection (HPAEC-PAD).

GH115 ␣-Glucuronidase
connecting ␤-1 and ␤-2 contains a pair of anti-parallel ␤-strands, whereas there is a short ␣-helix in the loop connecting ␤-5 with ␤-6. The two protomers in BoAgu115A display a "butterfly" like structure in which the two subunits make several interactions through the helical bundle domains. In addition, the C-terminal ␤-sandwich domain of protomer-1 makes extensive interactions with the TIM-barrel domain of protomer-2 and vice versa (Fig. 5, B and C). Identifying the Active Site of BoAgu115A-Inspection of the surface of the BoAgu115A protomers revealed a cleft-like structure in the (␤/␣) 8 domain that extends over the central ␤-barrel, on the opposite surface to the N-terminal domain. The center of the cleft, which may accommodate the xylan backbone, abuts onto a deep pocket that likely comprises the active site. This hypothesis was confirmed when BoAgu115A was co-crystallized with GlcA, which revealed electron density for the uronic acid in the proposed active site pocket (Figs. 5, B and C, 6, and  7). The sugar was found to be in its ring open conformation, which is highly unusual for hexaose aldehyde sugars. It is possible that the specific interactions between the uronic acid and the enzyme encourage GlcA to adopt an open chain conformation after glycosidic bond cleavage. The direct interactions between GlcA and the enzyme are as follows (Fig. 6): O3 makes a polar contact with N⑀-1 of Trp-249. O6A and O6B of the uronic acid carboxylate form polar contacts with the OH of Tyr-420 and N⑀-2 of His-422, respectively, the endocyclic oxygen interacts with the N of Val-426, and O1 makes a hydrogen bond with Asp-206. O4 is pointing into a hydrophobic open channel, comprising the side chains of Phe-203, Met-248, and Trp-245, which likely make apolar contacts with the methyl group of MeGlcA. Unlike the vast majority of glycoside hydrolases, the active site of BoAgu115A does not make typical parallel stacking interactions between the planar faces of aromatic residues and the sugar ring, although edge on edge apolar interactions between the sugar and two aromatic residues, Trp-249 and Tyr-420, and the aliphatic side chain of Val-426 were evident. The O␦-2 of Asp-332 in the apo structure (the amino acid is not visible in the structure of the BoAgu115A-GlcA complex, discussed in detail below), is in an appropriate position to make polar contacts with O2 of GlcA, whereas O␦-1 of Asp-206 forms a polar contact with O1 and may therefore act as the catalytic acid (Fig. 6).
To explore further the significance of the interactions between GlcA and BoAgu115A, and the function of other resi- dues in the active site of the enzyme, targeted amino acids were substituted with alanine and the activity of the enzyme variants determined. The data, presented in Table 3 and Fig. 6, showed that mutating the residues that interact with the carboxylate of GlcA, His-422 and Tyr-420, resulted in a ϳ10 5 -and ϳ10 3 -fold reduction in activity, respectively, indicating that this component of the substrate is an important specificity determinant. O3 appears to interact with Trp-249 as the W249A mutation caused a ϳ3000-fold reduction in catalytic activity. In addition to the residues that interact directly with the bound GlcA, several other amino acids in, or within the vicinity of, the active site were shown to make a significant contribution to the activity of the enzyme; the putative function of these residues is discussed below. Substituting amino acids near the putative xylan binding cleft/active site generally caused a modest reduction (5-100-fold) in the activity of the glucuronidase (see below), and mutating acid residues located in the other pocket/cleft-like structure on the enzyme (E162A, W169A, D192A, and D478A) did not affect activity (Table 3). These data are consistent with the view that the pocket located in the cleft-like structure in the TIM-barrel domain comprises the active site of BoAgu115A. The Possible Assembly of the Catalytic Apparatus of BoAgu115A-GH115 enzymes are inverting glycoside hydrolase (14) and thus glycosidic bond cleavage is catalyzed by an acid base-assisted single displacement mechanism. In such a mechanism the catalytic Brønsted base, typically a carboxylate, activates a water molecule that attacks the anomeric carbon of the glycone sugar, whereas the catalytic acid, also a carboxylic acid residue, donates a proton to the glycosidic oxygen promoting leaving group departure (14). As stated above, Asp-206, based on its interaction with O1 of GlcA, may comprise the catalytic acid. Support for this view is provided by the apolar environment of the aspartate; the carboxylate amino acid is in close proximity with Trp-249 and Val-426, promoting an elevated pK a . Furthermore, O␦-2 of Asp-206 is within hydrogen bonding distance with O␦-1 of Asp-424, and thus may function as the pK a modulator of the putative catalytic acid (Fig. 6). It should be emphasized, however, that the assignment of Asp-206 as the catalytic residue must be viewed with some caution as the D206A substitution resulted in only a 300-fold reduction in activity, and, as the bound GlcA is in an open chain form, the position of O1 may not reflect the location of the atom when the uronic acid is in a closed pyranose configuration, adopted in the substrate.
Based on mutagenesis data the other carboxylate amino acids in the vicinity of the active site, Asp-332 and Glu-375, are candidate catalytic residues; the D332A and E375A mutations caused complete inactivation of the enzyme and a ϳ5000-fold reduction in activity, respectively (Table 3). It is evident, how-ever, that the distance between Asp-332 and Glu-375 (in the apo structure) and the O1 of GlcA (in the structure of BoAgu115A in complex with GlcA); 5.4 Å and 11.8 Å, respectively, are greater than the normal ϳ3.5 Å between the catalytic base and the anomeric carbon of the glycone sugar in inverting glycoside hydrolases (Fig. 6). It is possible that BoAgu115A displays a "Grotthus"-style mechanism in which a remote amino acid activates the active site nucleophilic water via a string of solvent molecules (37), as proposed for some inverting glycoside hydrolases (37,38). In such a mechanism the distance between the catalytic base and the sugar is not restricted to 3.5 Å. Both Asp-332 and Glu-375, however, are on highly mobile loops that have high B-factors. Indeed, in the BoAgu115A-GlcA structure the loop that contains Asp-332 (Met-327 to Asn-346) was too disordered to be modeled, whereas Glu-375 could not be built past the ␤-carbon (Fig. 6). Thus, as discussed below, in the Michaelis catalytic complex Glu-375 and Asp-332 may adopt conformations, not observed in the apo and product structures, which reveal their true function in substrate binding and catalysis.
In enzymes and non-catalytic binding proteins that interact with glucuronic acid, arginines are often important specificity determinates, making bidentate hydrogen bonds/salt bridges with the carboxylate group (9,39). It is possible that in BoAgu115A Arg-328 contributes to substrate binding by also interacting with the carboxylate of the glucuronic acid substrate, a view supported by the observation that the R328A mutation caused ϳ10 5 -fold decrease in activity (Table 3). Arg-328 (and Glu-375) in the apo structure, however, are too distant from the hub of the active site to play a direct role in enzyme function (Fig. 6). Thus, for the arginine to contribute to substrate binding the loop containing this amino acid (and Asp-332) would be required to undergo a conformational change. Another possibility is that the role of Arg-328 is to stabilize the conformation of Asp-332, which may function as the catalytic base, whereas the primary role of Glu-375 could be to orientate the guanidino group of the arginine toward the aspartate. To conclude, the above structural and mutagenesis data indicate that the assembly of the Michaelis complex in BoAgu115A requires the repositioning of several flexible elements, most notably the loop carrying Arg-328 and Asp-332. Indeed, close inspection of the apo structure reveals that the conformation of this loop, which forms a component of the xylan binding cleft, in addition to part of the active site pocket, likely plays a central role in both glucuronoxylan binding and departure of the reaction products, both GlcA and the undecorated xylose polymer (Fig. 7, A and B). Furthermore, the loop also plays a role in stabilizing residues in the putative xylan binding cleft, notably Lys-374 and Glu-782, whose side chains are disordered in the GlcA bound structure where the loop is too flexible to be modeled (Fig. 7C). Although conformational changes are unusual in glycoside hydrolases, they are not without precedent, exemplified by the substantial movement of the catalytic TIM-barrel domain induced by substrate binding observed in GH112 glycoside phosphorylases (40).
Whether Asp-332, or possibly Glu-375, comprise the actual catalytic base is currently unclear, as is the role of Arg-328 in substrate binding, as a Michaelis complex could not be a Wild type and mutants of BoAgu115A were assayed using 1 mM XUXX as the substrate in 20 mM sodium phosphate buffer, pH 7.0, containing 1 mg ml Ϫ1 of BSA. At this substrate concentration the initial rate provides a direct readout of the catalytic efficiency of the enzyme. b ϳ780; the activity of the mutants of residues from the other cleft/pocket-like structure on the enzyme was estimated from a single time point reaction. c ND, no activity could be determined using an assay that can detect activity that is ϳ10 5 -fold less than the wild type enzyme.
obtained. Although the catalytic apparatus of glycoside hydrolases generally comprise two carboxylate residues, exceptions are evident with histidine acting as a catalytic acid-base in a GH3 N-acetylglucosaminidase (41), and as a catalytic acid in an inverting GH117 anhydro-L-galactosidase (42). Given that alanine substitution of His-422 caused a ϳ10 5 -fold reduction in activity, it is also possible that this residue is a component of the catalytic apparatus. Sequence alignments showed that the residues that had the greatest influence on BoAgu115A activity (mutants of amino acids that cause Ͼ10 4 -fold reduction in activity, those that interact with GlcA, and the proposed catalytic acid and pK a modulator, Asp-206 and Asp-424, respectively), are invariant in the other three GH115 enzymes that were shown to display ␣-glucuronidase activity (12,13,15) (Fig. 8). These data are entirely consistent with the proposed location of the BoAgu115A active site, and the likely identity of the amino acids that contribute to catalysis.
The Potential Role of the C-terminal Domain of BoAgu115A-Inspection of the dimeric form of BoAgu115A suggests that the C-terminal ␤-sandwich domain of protomer 2 contributes to one of the faces of the likely xylan binding cleft of protomer 1 and vice versa (Fig. 7). To explore the role of the C-terminal domain in the activity of the enzyme, alanine substitutions were made of Glu-782, Glu-785, Tyr-788, and Tyr-792, whose side chains line the proposed xylan binding cleft. The 100-fold reduction in catalytic activity mediated by the Y792A substitution (Table 3) indicated that Tyr-792 makes a significant contribution to xylan binding, whereas removal of the C-terminal domain (enzyme truncated at residue Ile-665) resulted in the complete abrogation of enzyme activity (Table 3), supporting a key role for this structural element in the topology of the GH115 catalytic apparatus.
Structural Comparison of GH115 and GH67 ␣-Glucuronidases-Structural comparison of BoAgu115A with the PDB database (using DaliLite version 3) identified the Geobacillus stearothermophilus GH67 ␣-glucuronidase, AguA (11), as the closest, albeit weak, structural homolog with a Z-score of 16.1, root mean square deviation of 4.7 Å, and sequence identity of 11% over 470 aligned resides. Despite the weak structural homology and sequence similarity, the first three domains of the two proteins display the same fold and, intriguingly, the location of the active site of the GH115 and GH67 enzymes, endo-and exo-acting xylan-specific ␣-glucuronidase, respectively, is conserved, pointing to a distant evolutionary link between the two families (Fig. 9A). It is evident, however, that there is no conservation of the active site residues, although it should be noted that the catalytic acid in GH67 enzymes is positioned on a highly mobile loop that can only be stabilized in the presence of ligands (11), which has some resonance with the mobile loop containing the catalytically significant residues Arg-328 and Asp-332 in BoAgu115A (Fig. 9B). In addition, the topology of the active site of the two enzymes is very different, comprising a deep pocket embedded in a rigid blind canyon (GH67 AguA) or a more accessible binding cleft (BoAgu115A), consistent with the exo and endo specificities of the respective glucuronidases. In general the catalytic apparatus is conserved in enzymes that are structurally related and display the same substrate specificity, but exhibit distinct modes of action; the introduction of exo activity is mediated by the modification of the substrate binding cleft of endo-acting enzymes such that the cleft is converted into a blind canyon or a tunnel (reviewed by Refs. 5 and 6). It is intriguing, therefore, that despite significant structural conservation, indicating an evolutionary link between the two glucuronidase families, the region of the enzymes that are most likely to be highly conserved, the cata- lytic apparatus, is clearly different in GH67 and GH115. The evolutionary selection pressures that led to such diversity in the active site of structurally and functionally related enzymes families are currently unclear.
Conclusions-This report reveals the first structure of a GH115 enzyme, identifying a likely distant evolutionary link to GH67, the other major family of ␣-glucuronidases. The locations of the residues that contribute to activity indicate that the enzyme undergoes a substantial conformational change to assemble a functional catalytic apparatus. It is interesting to note that despite complete conservation of the catalytically important amino acids in seven GH115 proteins, only BoAgu115A was shown to display activity (against glucuronoxylan). Although it is possible that the other six B. ovatus GH115 proteins are not catalytically active, a more likely explanation is that their target uronic acid-containing substrates are not glucuronoxylan, suggesting that GH115 is a poly-specific family. with AguA, a GH67 ␣-glucuronidase (magenta; PDB code 1MQQ) from G. stearothermophilus (11). The position of the bound GlcA from BoAgu115A and AguA are circled. Panel B shows an overlay of the active site of the two enzymes depicted in panel A in schematic format with the key catalytic and sugar binding residues displayed as sticks. The residues in which the carbons are colored green (apo form of BoAgu115A; see Fig. 6 for the rationale for showing residues from the two structures) are from the GH115 enzyme and amino acids with carbons shown in magenta from the GH67 glucuronidase. The carbons of the GlcA in complex with BoAgu115A and the GlcA bound to AguA are shown in salmon pink and slate blue, respectively, in both panels. All oxygens and nitrogens are shown in red and dark blue, respectively.